Chinese AI models are surpassing their US counterparts in cryptocurrency trading, according to blockchain analytics platform CoinGlass, as competition among leading generative AI chatbots escalates.
AI chatbots DeepSeek and Qwen3 Max, both developed in China, excelled in the ongoing crypto trading experiment on Wednesday, with DeepSeek being the sole AI model to achieve a positive unrealized return of 9.1%.
Qwen3, an AI model from Alibaba Cloud, followed in second place with a 0.5% unrealized loss, and Grok came next with a 1.24% unrealized loss, according to CoinGlass.
OpenAI’s ChatGPT-5 fell to the last position, recording over a 66% loss, decreasing its initial account value from $10,000 to just $3,453 at the time of this report.
The results have caught crypto traders off guard, considering that DeepSeek was developed at a significantly lower cost compared to its US competitors.
DeepSeek’s success was attributed to its bullish strategies on the crypto market, as it took leveraged long positions across major cryptocurrencies such as Bitcoin (BTC), Ether (ETH), Solana (SOL), BNB (BNB), Dogecoin (DOGE), and XRP (XRP).
Related: Arthur Hayes calls for $1M Bitcoin as new Japan PM orders economic stimulus
DeepSeek exceeds all AI models with just $5.3 million in training costs
DeepSeek was created with a total training cost of $5.3 million, according to its technical documentation.
In contrast, OpenAI has achieved a $500 billion valuation, establishing itself as the world’s largest startup, as reported by Cointelegraph on October 2. The company has amassed $57 billion in capital across 11 funding rounds, according to the company database platform Tracxn.
While precise figures on ChatGPT-5’s training budget remain undisclosed, OpenAI reportedly invested $5.7 billion in research and development in the first half of 2025, as highlighted by Reuters in September.
Estimates suggest that the total training cost for ChatGPT-5 ranges from $1.7 billion to $2.5 billion, according to a May 2024 post by chartered financial analyst Vladimir Kiselev.
Related: $19B market crash paves way for Bitcoin’s rise to $200K: Standard Chartered
AI models’ trading discrepancies may result from training data, says Nansen analyst
The disparity in crypto trading performance among the AI models is likely due to differences in their training data, according to Nicolai Sondergaard, a research analyst at crypto intelligence platform Nansen.
While ChatGPT is an excellent “general-purpose” large language model (LLM), Claude — another AI model — is primarily utilized for coding, the analyst informed Cointelegraph, adding:
“Looking over the historical PNLs so far, models generally have very large price swings, like being up $3,000 – $4,000 but then making a bad trade or getting caught on big moves, causing the LLM to close the trade.”
The effectiveness of some of these AI models could be heightened with tailored prompts, especially for ChatGPT and Google’s Gemini, according to strategic adviser and former quantitative trader, Kasper Vandeloock.
“Perhaps ChatGPT & Gemini could perform better with a different prompt; LLMs thrive on the prompt, so maybe their default behaviors are suboptimal,” Vandeloock mentioned to Cointelegraph.
Although AI tools can assist in detecting market trend changes for day traders using social media and technical signals, traders cannot yet depend on them for autonomous trading.
The competition was initiated with a $200 starting capital for each bot, later increasing to $10,000 per model, with trades executed on the decentralized exchange Hyperliquid.
Magazine: Crypto traders ‘fool themselves’ with price predictions — Peter Brandt
